Hilariously, when I asked Google about this, it responded with the usually discourse about the lack of attribution in LLM results, despite itself offering links for every assertion it makes.

Really shows the disconnect between training data and output!